sysctl: simplify ->strategy
[linux-2.6/libata-dev.git] / kernel / sysctl.c
blob9792c315676afab76e7896dab7dead7d21a900a2
1 /*
2 * sysctl.c: General linux system control interface
4 * Begun 24 March 1995, Stephen Tweedie
5 * Added /proc support, Dec 1995
6 * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
7 * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
8 * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
9 * Dynamic registration fixes, Stephen Tweedie.
10 * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
11 * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
12 * Horn.
13 * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
14 * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
15 * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
16 * Wendling.
17 * The list_for_each() macro wasn't appropriate for the sysctl loop.
18 * Removed it and replaced it with older style, 03/23/00, Bill Wendling
21 #include <linux/module.h>
22 #include <linux/mm.h>
23 #include <linux/swap.h>
24 #include <linux/slab.h>
25 #include <linux/sysctl.h>
26 #include <linux/proc_fs.h>
27 #include <linux/security.h>
28 #include <linux/ctype.h>
29 #include <linux/utsname.h>
30 #include <linux/smp_lock.h>
31 #include <linux/fs.h>
32 #include <linux/init.h>
33 #include <linux/kernel.h>
34 #include <linux/kobject.h>
35 #include <linux/net.h>
36 #include <linux/sysrq.h>
37 #include <linux/highuid.h>
38 #include <linux/writeback.h>
39 #include <linux/hugetlb.h>
40 #include <linux/initrd.h>
41 #include <linux/key.h>
42 #include <linux/times.h>
43 #include <linux/limits.h>
44 #include <linux/dcache.h>
45 #include <linux/syscalls.h>
46 #include <linux/vmstat.h>
47 #include <linux/nfs_fs.h>
48 #include <linux/acpi.h>
49 #include <linux/reboot.h>
50 #include <linux/ftrace.h>
52 #include <asm/uaccess.h>
53 #include <asm/processor.h>
55 #ifdef CONFIG_X86
56 #include <asm/nmi.h>
57 #include <asm/stacktrace.h>
58 #include <asm/io.h>
59 #endif
61 static int deprecated_sysctl_warning(struct __sysctl_args *args);
63 #if defined(CONFIG_SYSCTL)
65 /* External variables not in a header file. */
66 extern int C_A_D;
67 extern int print_fatal_signals;
68 extern int sysctl_overcommit_memory;
69 extern int sysctl_overcommit_ratio;
70 extern int sysctl_panic_on_oom;
71 extern int sysctl_oom_kill_allocating_task;
72 extern int sysctl_oom_dump_tasks;
73 extern int max_threads;
74 extern int core_uses_pid;
75 extern int suid_dumpable;
76 extern char core_pattern[];
77 extern int pid_max;
78 extern int min_free_kbytes;
79 extern int pid_max_min, pid_max_max;
80 extern int sysctl_drop_caches;
81 extern int percpu_pagelist_fraction;
82 extern int compat_log;
83 extern int latencytop_enabled;
84 extern int sysctl_nr_open_min, sysctl_nr_open_max;
85 #ifdef CONFIG_RCU_TORTURE_TEST
86 extern int rcutorture_runnable;
87 #endif /* #ifdef CONFIG_RCU_TORTURE_TEST */
89 /* Constants used for minimum and maximum */
90 #if defined(CONFIG_HIGHMEM) || defined(CONFIG_DETECT_SOFTLOCKUP)
91 static int one = 1;
92 #endif
94 #ifdef CONFIG_DETECT_SOFTLOCKUP
95 static int sixty = 60;
96 static int neg_one = -1;
97 #endif
99 #if defined(CONFIG_MMU) && defined(CONFIG_FILE_LOCKING)
100 static int two = 2;
101 #endif
103 static int zero;
104 static int one_hundred = 100;
106 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
107 static int maxolduid = 65535;
108 static int minolduid;
109 static int min_percpu_pagelist_fract = 8;
111 static int ngroups_max = NGROUPS_MAX;
113 #ifdef CONFIG_MODULES
114 extern char modprobe_path[];
115 #endif
116 #ifdef CONFIG_CHR_DEV_SG
117 extern int sg_big_buff;
118 #endif
120 #ifdef CONFIG_SPARC
121 #include <asm/system.h>
122 #endif
124 #ifdef __hppa__
125 extern int pwrsw_enabled;
126 extern int unaligned_enabled;
127 #endif
129 #ifdef CONFIG_S390
130 #ifdef CONFIG_MATHEMU
131 extern int sysctl_ieee_emulation_warnings;
132 #endif
133 extern int sysctl_userprocess_debug;
134 extern int spin_retry;
135 #endif
137 #ifdef CONFIG_BSD_PROCESS_ACCT
138 extern int acct_parm[];
139 #endif
141 #ifdef CONFIG_IA64
142 extern int no_unaligned_warning;
143 #endif
145 #ifdef CONFIG_RT_MUTEXES
146 extern int max_lock_depth;
147 #endif
149 #ifdef CONFIG_PROC_SYSCTL
150 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
151 void __user *buffer, size_t *lenp, loff_t *ppos);
152 static int proc_taint(struct ctl_table *table, int write, struct file *filp,
153 void __user *buffer, size_t *lenp, loff_t *ppos);
154 #endif
156 static struct ctl_table root_table[];
157 static struct ctl_table_root sysctl_table_root;
158 static struct ctl_table_header root_table_header = {
159 .count = 1,
160 .ctl_table = root_table,
161 .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
162 .root = &sysctl_table_root,
163 .set = &sysctl_table_root.default_set,
165 static struct ctl_table_root sysctl_table_root = {
166 .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
167 .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
170 static struct ctl_table kern_table[];
171 static struct ctl_table vm_table[];
172 static struct ctl_table fs_table[];
173 static struct ctl_table debug_table[];
174 static struct ctl_table dev_table[];
175 extern struct ctl_table random_table[];
176 #ifdef CONFIG_INOTIFY_USER
177 extern struct ctl_table inotify_table[];
178 #endif
180 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
181 int sysctl_legacy_va_layout;
182 #endif
184 extern int prove_locking;
185 extern int lock_stat;
187 /* The default sysctl tables: */
189 static struct ctl_table root_table[] = {
191 .ctl_name = CTL_KERN,
192 .procname = "kernel",
193 .mode = 0555,
194 .child = kern_table,
197 .ctl_name = CTL_VM,
198 .procname = "vm",
199 .mode = 0555,
200 .child = vm_table,
203 .ctl_name = CTL_FS,
204 .procname = "fs",
205 .mode = 0555,
206 .child = fs_table,
209 .ctl_name = CTL_DEBUG,
210 .procname = "debug",
211 .mode = 0555,
212 .child = debug_table,
215 .ctl_name = CTL_DEV,
216 .procname = "dev",
217 .mode = 0555,
218 .child = dev_table,
221 * NOTE: do not add new entries to this table unless you have read
222 * Documentation/sysctl/ctl_unnumbered.txt
224 { .ctl_name = 0 }
227 #ifdef CONFIG_SCHED_DEBUG
228 static int min_sched_granularity_ns = 100000; /* 100 usecs */
229 static int max_sched_granularity_ns = NSEC_PER_SEC; /* 1 second */
230 static int min_wakeup_granularity_ns; /* 0 usecs */
231 static int max_wakeup_granularity_ns = NSEC_PER_SEC; /* 1 second */
232 #endif
234 static struct ctl_table kern_table[] = {
235 #ifdef CONFIG_SCHED_DEBUG
237 .ctl_name = CTL_UNNUMBERED,
238 .procname = "sched_min_granularity_ns",
239 .data = &sysctl_sched_min_granularity,
240 .maxlen = sizeof(unsigned int),
241 .mode = 0644,
242 .proc_handler = &sched_nr_latency_handler,
243 .strategy = &sysctl_intvec,
244 .extra1 = &min_sched_granularity_ns,
245 .extra2 = &max_sched_granularity_ns,
248 .ctl_name = CTL_UNNUMBERED,
249 .procname = "sched_latency_ns",
250 .data = &sysctl_sched_latency,
251 .maxlen = sizeof(unsigned int),
252 .mode = 0644,
253 .proc_handler = &sched_nr_latency_handler,
254 .strategy = &sysctl_intvec,
255 .extra1 = &min_sched_granularity_ns,
256 .extra2 = &max_sched_granularity_ns,
259 .ctl_name = CTL_UNNUMBERED,
260 .procname = "sched_wakeup_granularity_ns",
261 .data = &sysctl_sched_wakeup_granularity,
262 .maxlen = sizeof(unsigned int),
263 .mode = 0644,
264 .proc_handler = &proc_dointvec_minmax,
265 .strategy = &sysctl_intvec,
266 .extra1 = &min_wakeup_granularity_ns,
267 .extra2 = &max_wakeup_granularity_ns,
270 .ctl_name = CTL_UNNUMBERED,
271 .procname = "sched_shares_ratelimit",
272 .data = &sysctl_sched_shares_ratelimit,
273 .maxlen = sizeof(unsigned int),
274 .mode = 0644,
275 .proc_handler = &proc_dointvec,
278 .ctl_name = CTL_UNNUMBERED,
279 .procname = "sched_child_runs_first",
280 .data = &sysctl_sched_child_runs_first,
281 .maxlen = sizeof(unsigned int),
282 .mode = 0644,
283 .proc_handler = &proc_dointvec,
286 .ctl_name = CTL_UNNUMBERED,
287 .procname = "sched_features",
288 .data = &sysctl_sched_features,
289 .maxlen = sizeof(unsigned int),
290 .mode = 0644,
291 .proc_handler = &proc_dointvec,
294 .ctl_name = CTL_UNNUMBERED,
295 .procname = "sched_migration_cost",
296 .data = &sysctl_sched_migration_cost,
297 .maxlen = sizeof(unsigned int),
298 .mode = 0644,
299 .proc_handler = &proc_dointvec,
302 .ctl_name = CTL_UNNUMBERED,
303 .procname = "sched_nr_migrate",
304 .data = &sysctl_sched_nr_migrate,
305 .maxlen = sizeof(unsigned int),
306 .mode = 0644,
307 .proc_handler = &proc_dointvec,
309 #endif
311 .ctl_name = CTL_UNNUMBERED,
312 .procname = "sched_rt_period_us",
313 .data = &sysctl_sched_rt_period,
314 .maxlen = sizeof(unsigned int),
315 .mode = 0644,
316 .proc_handler = &sched_rt_handler,
319 .ctl_name = CTL_UNNUMBERED,
320 .procname = "sched_rt_runtime_us",
321 .data = &sysctl_sched_rt_runtime,
322 .maxlen = sizeof(int),
323 .mode = 0644,
324 .proc_handler = &sched_rt_handler,
327 .ctl_name = CTL_UNNUMBERED,
328 .procname = "sched_compat_yield",
329 .data = &sysctl_sched_compat_yield,
330 .maxlen = sizeof(unsigned int),
331 .mode = 0644,
332 .proc_handler = &proc_dointvec,
334 #ifdef CONFIG_PROVE_LOCKING
336 .ctl_name = CTL_UNNUMBERED,
337 .procname = "prove_locking",
338 .data = &prove_locking,
339 .maxlen = sizeof(int),
340 .mode = 0644,
341 .proc_handler = &proc_dointvec,
343 #endif
344 #ifdef CONFIG_LOCK_STAT
346 .ctl_name = CTL_UNNUMBERED,
347 .procname = "lock_stat",
348 .data = &lock_stat,
349 .maxlen = sizeof(int),
350 .mode = 0644,
351 .proc_handler = &proc_dointvec,
353 #endif
355 .ctl_name = KERN_PANIC,
356 .procname = "panic",
357 .data = &panic_timeout,
358 .maxlen = sizeof(int),
359 .mode = 0644,
360 .proc_handler = &proc_dointvec,
363 .ctl_name = KERN_CORE_USES_PID,
364 .procname = "core_uses_pid",
365 .data = &core_uses_pid,
366 .maxlen = sizeof(int),
367 .mode = 0644,
368 .proc_handler = &proc_dointvec,
371 .ctl_name = KERN_CORE_PATTERN,
372 .procname = "core_pattern",
373 .data = core_pattern,
374 .maxlen = CORENAME_MAX_SIZE,
375 .mode = 0644,
376 .proc_handler = &proc_dostring,
377 .strategy = &sysctl_string,
379 #ifdef CONFIG_PROC_SYSCTL
381 .procname = "tainted",
382 .maxlen = sizeof(long),
383 .mode = 0644,
384 .proc_handler = &proc_taint,
386 #endif
387 #ifdef CONFIG_LATENCYTOP
389 .procname = "latencytop",
390 .data = &latencytop_enabled,
391 .maxlen = sizeof(int),
392 .mode = 0644,
393 .proc_handler = &proc_dointvec,
395 #endif
396 #ifdef CONFIG_BLK_DEV_INITRD
398 .ctl_name = KERN_REALROOTDEV,
399 .procname = "real-root-dev",
400 .data = &real_root_dev,
401 .maxlen = sizeof(int),
402 .mode = 0644,
403 .proc_handler = &proc_dointvec,
405 #endif
407 .ctl_name = CTL_UNNUMBERED,
408 .procname = "print-fatal-signals",
409 .data = &print_fatal_signals,
410 .maxlen = sizeof(int),
411 .mode = 0644,
412 .proc_handler = &proc_dointvec,
414 #ifdef CONFIG_SPARC
416 .ctl_name = KERN_SPARC_REBOOT,
417 .procname = "reboot-cmd",
418 .data = reboot_command,
419 .maxlen = 256,
420 .mode = 0644,
421 .proc_handler = &proc_dostring,
422 .strategy = &sysctl_string,
425 .ctl_name = KERN_SPARC_STOP_A,
426 .procname = "stop-a",
427 .data = &stop_a_enabled,
428 .maxlen = sizeof (int),
429 .mode = 0644,
430 .proc_handler = &proc_dointvec,
433 .ctl_name = KERN_SPARC_SCONS_PWROFF,
434 .procname = "scons-poweroff",
435 .data = &scons_pwroff,
436 .maxlen = sizeof (int),
437 .mode = 0644,
438 .proc_handler = &proc_dointvec,
440 #endif
441 #ifdef __hppa__
443 .ctl_name = KERN_HPPA_PWRSW,
444 .procname = "soft-power",
445 .data = &pwrsw_enabled,
446 .maxlen = sizeof (int),
447 .mode = 0644,
448 .proc_handler = &proc_dointvec,
451 .ctl_name = KERN_HPPA_UNALIGNED,
452 .procname = "unaligned-trap",
453 .data = &unaligned_enabled,
454 .maxlen = sizeof (int),
455 .mode = 0644,
456 .proc_handler = &proc_dointvec,
458 #endif
460 .ctl_name = KERN_CTLALTDEL,
461 .procname = "ctrl-alt-del",
462 .data = &C_A_D,
463 .maxlen = sizeof(int),
464 .mode = 0644,
465 .proc_handler = &proc_dointvec,
467 #ifdef CONFIG_FTRACE
469 .ctl_name = CTL_UNNUMBERED,
470 .procname = "ftrace_enabled",
471 .data = &ftrace_enabled,
472 .maxlen = sizeof(int),
473 .mode = 0644,
474 .proc_handler = &ftrace_enable_sysctl,
476 #endif
477 #ifdef CONFIG_MODULES
479 .ctl_name = KERN_MODPROBE,
480 .procname = "modprobe",
481 .data = &modprobe_path,
482 .maxlen = KMOD_PATH_LEN,
483 .mode = 0644,
484 .proc_handler = &proc_dostring,
485 .strategy = &sysctl_string,
487 #endif
488 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
490 .ctl_name = KERN_HOTPLUG,
491 .procname = "hotplug",
492 .data = &uevent_helper,
493 .maxlen = UEVENT_HELPER_PATH_LEN,
494 .mode = 0644,
495 .proc_handler = &proc_dostring,
496 .strategy = &sysctl_string,
498 #endif
499 #ifdef CONFIG_CHR_DEV_SG
501 .ctl_name = KERN_SG_BIG_BUFF,
502 .procname = "sg-big-buff",
503 .data = &sg_big_buff,
504 .maxlen = sizeof (int),
505 .mode = 0444,
506 .proc_handler = &proc_dointvec,
508 #endif
509 #ifdef CONFIG_BSD_PROCESS_ACCT
511 .ctl_name = KERN_ACCT,
512 .procname = "acct",
513 .data = &acct_parm,
514 .maxlen = 3*sizeof(int),
515 .mode = 0644,
516 .proc_handler = &proc_dointvec,
518 #endif
519 #ifdef CONFIG_MAGIC_SYSRQ
521 .ctl_name = KERN_SYSRQ,
522 .procname = "sysrq",
523 .data = &__sysrq_enabled,
524 .maxlen = sizeof (int),
525 .mode = 0644,
526 .proc_handler = &proc_dointvec,
528 #endif
529 #ifdef CONFIG_PROC_SYSCTL
531 .procname = "cad_pid",
532 .data = NULL,
533 .maxlen = sizeof (int),
534 .mode = 0600,
535 .proc_handler = &proc_do_cad_pid,
537 #endif
539 .ctl_name = KERN_MAX_THREADS,
540 .procname = "threads-max",
541 .data = &max_threads,
542 .maxlen = sizeof(int),
543 .mode = 0644,
544 .proc_handler = &proc_dointvec,
547 .ctl_name = KERN_RANDOM,
548 .procname = "random",
549 .mode = 0555,
550 .child = random_table,
553 .ctl_name = KERN_OVERFLOWUID,
554 .procname = "overflowuid",
555 .data = &overflowuid,
556 .maxlen = sizeof(int),
557 .mode = 0644,
558 .proc_handler = &proc_dointvec_minmax,
559 .strategy = &sysctl_intvec,
560 .extra1 = &minolduid,
561 .extra2 = &maxolduid,
564 .ctl_name = KERN_OVERFLOWGID,
565 .procname = "overflowgid",
566 .data = &overflowgid,
567 .maxlen = sizeof(int),
568 .mode = 0644,
569 .proc_handler = &proc_dointvec_minmax,
570 .strategy = &sysctl_intvec,
571 .extra1 = &minolduid,
572 .extra2 = &maxolduid,
574 #ifdef CONFIG_S390
575 #ifdef CONFIG_MATHEMU
577 .ctl_name = KERN_IEEE_EMULATION_WARNINGS,
578 .procname = "ieee_emulation_warnings",
579 .data = &sysctl_ieee_emulation_warnings,
580 .maxlen = sizeof(int),
581 .mode = 0644,
582 .proc_handler = &proc_dointvec,
584 #endif
586 .ctl_name = KERN_S390_USER_DEBUG_LOGGING,
587 .procname = "userprocess_debug",
588 .data = &sysctl_userprocess_debug,
589 .maxlen = sizeof(int),
590 .mode = 0644,
591 .proc_handler = &proc_dointvec,
593 #endif
595 .ctl_name = KERN_PIDMAX,
596 .procname = "pid_max",
597 .data = &pid_max,
598 .maxlen = sizeof (int),
599 .mode = 0644,
600 .proc_handler = &proc_dointvec_minmax,
601 .strategy = sysctl_intvec,
602 .extra1 = &pid_max_min,
603 .extra2 = &pid_max_max,
606 .ctl_name = KERN_PANIC_ON_OOPS,
607 .procname = "panic_on_oops",
608 .data = &panic_on_oops,
609 .maxlen = sizeof(int),
610 .mode = 0644,
611 .proc_handler = &proc_dointvec,
613 #if defined CONFIG_PRINTK
615 .ctl_name = KERN_PRINTK,
616 .procname = "printk",
617 .data = &console_loglevel,
618 .maxlen = 4*sizeof(int),
619 .mode = 0644,
620 .proc_handler = &proc_dointvec,
623 .ctl_name = KERN_PRINTK_RATELIMIT,
624 .procname = "printk_ratelimit",
625 .data = &printk_ratelimit_state.interval,
626 .maxlen = sizeof(int),
627 .mode = 0644,
628 .proc_handler = &proc_dointvec_jiffies,
629 .strategy = &sysctl_jiffies,
632 .ctl_name = KERN_PRINTK_RATELIMIT_BURST,
633 .procname = "printk_ratelimit_burst",
634 .data = &printk_ratelimit_state.burst,
635 .maxlen = sizeof(int),
636 .mode = 0644,
637 .proc_handler = &proc_dointvec,
639 #endif
641 .ctl_name = KERN_NGROUPS_MAX,
642 .procname = "ngroups_max",
643 .data = &ngroups_max,
644 .maxlen = sizeof (int),
645 .mode = 0444,
646 .proc_handler = &proc_dointvec,
648 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
650 .ctl_name = KERN_UNKNOWN_NMI_PANIC,
651 .procname = "unknown_nmi_panic",
652 .data = &unknown_nmi_panic,
653 .maxlen = sizeof (int),
654 .mode = 0644,
655 .proc_handler = &proc_dointvec,
658 .procname = "nmi_watchdog",
659 .data = &nmi_watchdog_enabled,
660 .maxlen = sizeof (int),
661 .mode = 0644,
662 .proc_handler = &proc_nmi_enabled,
664 #endif
665 #if defined(CONFIG_X86)
667 .ctl_name = KERN_PANIC_ON_NMI,
668 .procname = "panic_on_unrecovered_nmi",
669 .data = &panic_on_unrecovered_nmi,
670 .maxlen = sizeof(int),
671 .mode = 0644,
672 .proc_handler = &proc_dointvec,
675 .ctl_name = KERN_BOOTLOADER_TYPE,
676 .procname = "bootloader_type",
677 .data = &bootloader_type,
678 .maxlen = sizeof (int),
679 .mode = 0444,
680 .proc_handler = &proc_dointvec,
683 .ctl_name = CTL_UNNUMBERED,
684 .procname = "kstack_depth_to_print",
685 .data = &kstack_depth_to_print,
686 .maxlen = sizeof(int),
687 .mode = 0644,
688 .proc_handler = &proc_dointvec,
691 .ctl_name = CTL_UNNUMBERED,
692 .procname = "io_delay_type",
693 .data = &io_delay_type,
694 .maxlen = sizeof(int),
695 .mode = 0644,
696 .proc_handler = &proc_dointvec,
698 #endif
699 #if defined(CONFIG_MMU)
701 .ctl_name = KERN_RANDOMIZE,
702 .procname = "randomize_va_space",
703 .data = &randomize_va_space,
704 .maxlen = sizeof(int),
705 .mode = 0644,
706 .proc_handler = &proc_dointvec,
708 #endif
709 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
711 .ctl_name = KERN_SPIN_RETRY,
712 .procname = "spin_retry",
713 .data = &spin_retry,
714 .maxlen = sizeof (int),
715 .mode = 0644,
716 .proc_handler = &proc_dointvec,
718 #endif
719 #if defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
721 .procname = "acpi_video_flags",
722 .data = &acpi_realmode_flags,
723 .maxlen = sizeof (unsigned long),
724 .mode = 0644,
725 .proc_handler = &proc_doulongvec_minmax,
727 #endif
728 #ifdef CONFIG_IA64
730 .ctl_name = KERN_IA64_UNALIGNED,
731 .procname = "ignore-unaligned-usertrap",
732 .data = &no_unaligned_warning,
733 .maxlen = sizeof (int),
734 .mode = 0644,
735 .proc_handler = &proc_dointvec,
737 #endif
738 #ifdef CONFIG_DETECT_SOFTLOCKUP
740 .ctl_name = CTL_UNNUMBERED,
741 .procname = "softlockup_panic",
742 .data = &softlockup_panic,
743 .maxlen = sizeof(int),
744 .mode = 0644,
745 .proc_handler = &proc_dointvec_minmax,
746 .strategy = &sysctl_intvec,
747 .extra1 = &zero,
748 .extra2 = &one,
751 .ctl_name = CTL_UNNUMBERED,
752 .procname = "softlockup_thresh",
753 .data = &softlockup_thresh,
754 .maxlen = sizeof(int),
755 .mode = 0644,
756 .proc_handler = &proc_dointvec_minmax,
757 .strategy = &sysctl_intvec,
758 .extra1 = &neg_one,
759 .extra2 = &sixty,
762 .ctl_name = CTL_UNNUMBERED,
763 .procname = "hung_task_check_count",
764 .data = &sysctl_hung_task_check_count,
765 .maxlen = sizeof(unsigned long),
766 .mode = 0644,
767 .proc_handler = &proc_doulongvec_minmax,
768 .strategy = &sysctl_intvec,
771 .ctl_name = CTL_UNNUMBERED,
772 .procname = "hung_task_timeout_secs",
773 .data = &sysctl_hung_task_timeout_secs,
774 .maxlen = sizeof(unsigned long),
775 .mode = 0644,
776 .proc_handler = &proc_doulongvec_minmax,
777 .strategy = &sysctl_intvec,
780 .ctl_name = CTL_UNNUMBERED,
781 .procname = "hung_task_warnings",
782 .data = &sysctl_hung_task_warnings,
783 .maxlen = sizeof(unsigned long),
784 .mode = 0644,
785 .proc_handler = &proc_doulongvec_minmax,
786 .strategy = &sysctl_intvec,
788 #endif
789 #ifdef CONFIG_COMPAT
791 .ctl_name = KERN_COMPAT_LOG,
792 .procname = "compat-log",
793 .data = &compat_log,
794 .maxlen = sizeof (int),
795 .mode = 0644,
796 .proc_handler = &proc_dointvec,
798 #endif
799 #ifdef CONFIG_RT_MUTEXES
801 .ctl_name = KERN_MAX_LOCK_DEPTH,
802 .procname = "max_lock_depth",
803 .data = &max_lock_depth,
804 .maxlen = sizeof(int),
805 .mode = 0644,
806 .proc_handler = &proc_dointvec,
808 #endif
810 .ctl_name = CTL_UNNUMBERED,
811 .procname = "poweroff_cmd",
812 .data = &poweroff_cmd,
813 .maxlen = POWEROFF_CMD_PATH_LEN,
814 .mode = 0644,
815 .proc_handler = &proc_dostring,
816 .strategy = &sysctl_string,
818 #ifdef CONFIG_KEYS
820 .ctl_name = CTL_UNNUMBERED,
821 .procname = "keys",
822 .mode = 0555,
823 .child = key_sysctls,
825 #endif
826 #ifdef CONFIG_RCU_TORTURE_TEST
828 .ctl_name = CTL_UNNUMBERED,
829 .procname = "rcutorture_runnable",
830 .data = &rcutorture_runnable,
831 .maxlen = sizeof(int),
832 .mode = 0644,
833 .proc_handler = &proc_dointvec,
835 #endif
837 * NOTE: do not add new entries to this table unless you have read
838 * Documentation/sysctl/ctl_unnumbered.txt
840 { .ctl_name = 0 }
843 static struct ctl_table vm_table[] = {
845 .ctl_name = VM_OVERCOMMIT_MEMORY,
846 .procname = "overcommit_memory",
847 .data = &sysctl_overcommit_memory,
848 .maxlen = sizeof(sysctl_overcommit_memory),
849 .mode = 0644,
850 .proc_handler = &proc_dointvec,
853 .ctl_name = VM_PANIC_ON_OOM,
854 .procname = "panic_on_oom",
855 .data = &sysctl_panic_on_oom,
856 .maxlen = sizeof(sysctl_panic_on_oom),
857 .mode = 0644,
858 .proc_handler = &proc_dointvec,
861 .ctl_name = CTL_UNNUMBERED,
862 .procname = "oom_kill_allocating_task",
863 .data = &sysctl_oom_kill_allocating_task,
864 .maxlen = sizeof(sysctl_oom_kill_allocating_task),
865 .mode = 0644,
866 .proc_handler = &proc_dointvec,
869 .ctl_name = CTL_UNNUMBERED,
870 .procname = "oom_dump_tasks",
871 .data = &sysctl_oom_dump_tasks,
872 .maxlen = sizeof(sysctl_oom_dump_tasks),
873 .mode = 0644,
874 .proc_handler = &proc_dointvec,
877 .ctl_name = VM_OVERCOMMIT_RATIO,
878 .procname = "overcommit_ratio",
879 .data = &sysctl_overcommit_ratio,
880 .maxlen = sizeof(sysctl_overcommit_ratio),
881 .mode = 0644,
882 .proc_handler = &proc_dointvec,
885 .ctl_name = VM_PAGE_CLUSTER,
886 .procname = "page-cluster",
887 .data = &page_cluster,
888 .maxlen = sizeof(int),
889 .mode = 0644,
890 .proc_handler = &proc_dointvec,
893 .ctl_name = VM_DIRTY_BACKGROUND,
894 .procname = "dirty_background_ratio",
895 .data = &dirty_background_ratio,
896 .maxlen = sizeof(dirty_background_ratio),
897 .mode = 0644,
898 .proc_handler = &proc_dointvec_minmax,
899 .strategy = &sysctl_intvec,
900 .extra1 = &zero,
901 .extra2 = &one_hundred,
904 .ctl_name = VM_DIRTY_RATIO,
905 .procname = "dirty_ratio",
906 .data = &vm_dirty_ratio,
907 .maxlen = sizeof(vm_dirty_ratio),
908 .mode = 0644,
909 .proc_handler = &dirty_ratio_handler,
910 .strategy = &sysctl_intvec,
911 .extra1 = &zero,
912 .extra2 = &one_hundred,
915 .procname = "dirty_writeback_centisecs",
916 .data = &dirty_writeback_interval,
917 .maxlen = sizeof(dirty_writeback_interval),
918 .mode = 0644,
919 .proc_handler = &dirty_writeback_centisecs_handler,
922 .procname = "dirty_expire_centisecs",
923 .data = &dirty_expire_interval,
924 .maxlen = sizeof(dirty_expire_interval),
925 .mode = 0644,
926 .proc_handler = &proc_dointvec_userhz_jiffies,
929 .ctl_name = VM_NR_PDFLUSH_THREADS,
930 .procname = "nr_pdflush_threads",
931 .data = &nr_pdflush_threads,
932 .maxlen = sizeof nr_pdflush_threads,
933 .mode = 0444 /* read-only*/,
934 .proc_handler = &proc_dointvec,
937 .ctl_name = VM_SWAPPINESS,
938 .procname = "swappiness",
939 .data = &vm_swappiness,
940 .maxlen = sizeof(vm_swappiness),
941 .mode = 0644,
942 .proc_handler = &proc_dointvec_minmax,
943 .strategy = &sysctl_intvec,
944 .extra1 = &zero,
945 .extra2 = &one_hundred,
947 #ifdef CONFIG_HUGETLB_PAGE
949 .procname = "nr_hugepages",
950 .data = NULL,
951 .maxlen = sizeof(unsigned long),
952 .mode = 0644,
953 .proc_handler = &hugetlb_sysctl_handler,
954 .extra1 = (void *)&hugetlb_zero,
955 .extra2 = (void *)&hugetlb_infinity,
958 .ctl_name = VM_HUGETLB_GROUP,
959 .procname = "hugetlb_shm_group",
960 .data = &sysctl_hugetlb_shm_group,
961 .maxlen = sizeof(gid_t),
962 .mode = 0644,
963 .proc_handler = &proc_dointvec,
966 .ctl_name = CTL_UNNUMBERED,
967 .procname = "hugepages_treat_as_movable",
968 .data = &hugepages_treat_as_movable,
969 .maxlen = sizeof(int),
970 .mode = 0644,
971 .proc_handler = &hugetlb_treat_movable_handler,
974 .ctl_name = CTL_UNNUMBERED,
975 .procname = "nr_overcommit_hugepages",
976 .data = NULL,
977 .maxlen = sizeof(unsigned long),
978 .mode = 0644,
979 .proc_handler = &hugetlb_overcommit_handler,
980 .extra1 = (void *)&hugetlb_zero,
981 .extra2 = (void *)&hugetlb_infinity,
983 #endif
985 .ctl_name = VM_LOWMEM_RESERVE_RATIO,
986 .procname = "lowmem_reserve_ratio",
987 .data = &sysctl_lowmem_reserve_ratio,
988 .maxlen = sizeof(sysctl_lowmem_reserve_ratio),
989 .mode = 0644,
990 .proc_handler = &lowmem_reserve_ratio_sysctl_handler,
991 .strategy = &sysctl_intvec,
994 .ctl_name = VM_DROP_PAGECACHE,
995 .procname = "drop_caches",
996 .data = &sysctl_drop_caches,
997 .maxlen = sizeof(int),
998 .mode = 0644,
999 .proc_handler = drop_caches_sysctl_handler,
1000 .strategy = &sysctl_intvec,
1003 .ctl_name = VM_MIN_FREE_KBYTES,
1004 .procname = "min_free_kbytes",
1005 .data = &min_free_kbytes,
1006 .maxlen = sizeof(min_free_kbytes),
1007 .mode = 0644,
1008 .proc_handler = &min_free_kbytes_sysctl_handler,
1009 .strategy = &sysctl_intvec,
1010 .extra1 = &zero,
1013 .ctl_name = VM_PERCPU_PAGELIST_FRACTION,
1014 .procname = "percpu_pagelist_fraction",
1015 .data = &percpu_pagelist_fraction,
1016 .maxlen = sizeof(percpu_pagelist_fraction),
1017 .mode = 0644,
1018 .proc_handler = &percpu_pagelist_fraction_sysctl_handler,
1019 .strategy = &sysctl_intvec,
1020 .extra1 = &min_percpu_pagelist_fract,
1022 #ifdef CONFIG_MMU
1024 .ctl_name = VM_MAX_MAP_COUNT,
1025 .procname = "max_map_count",
1026 .data = &sysctl_max_map_count,
1027 .maxlen = sizeof(sysctl_max_map_count),
1028 .mode = 0644,
1029 .proc_handler = &proc_dointvec
1031 #endif
1033 .ctl_name = VM_LAPTOP_MODE,
1034 .procname = "laptop_mode",
1035 .data = &laptop_mode,
1036 .maxlen = sizeof(laptop_mode),
1037 .mode = 0644,
1038 .proc_handler = &proc_dointvec_jiffies,
1039 .strategy = &sysctl_jiffies,
1042 .ctl_name = VM_BLOCK_DUMP,
1043 .procname = "block_dump",
1044 .data = &block_dump,
1045 .maxlen = sizeof(block_dump),
1046 .mode = 0644,
1047 .proc_handler = &proc_dointvec,
1048 .strategy = &sysctl_intvec,
1049 .extra1 = &zero,
1052 .ctl_name = VM_VFS_CACHE_PRESSURE,
1053 .procname = "vfs_cache_pressure",
1054 .data = &sysctl_vfs_cache_pressure,
1055 .maxlen = sizeof(sysctl_vfs_cache_pressure),
1056 .mode = 0644,
1057 .proc_handler = &proc_dointvec,
1058 .strategy = &sysctl_intvec,
1059 .extra1 = &zero,
1061 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1063 .ctl_name = VM_LEGACY_VA_LAYOUT,
1064 .procname = "legacy_va_layout",
1065 .data = &sysctl_legacy_va_layout,
1066 .maxlen = sizeof(sysctl_legacy_va_layout),
1067 .mode = 0644,
1068 .proc_handler = &proc_dointvec,
1069 .strategy = &sysctl_intvec,
1070 .extra1 = &zero,
1072 #endif
1073 #ifdef CONFIG_NUMA
1075 .ctl_name = VM_ZONE_RECLAIM_MODE,
1076 .procname = "zone_reclaim_mode",
1077 .data = &zone_reclaim_mode,
1078 .maxlen = sizeof(zone_reclaim_mode),
1079 .mode = 0644,
1080 .proc_handler = &proc_dointvec,
1081 .strategy = &sysctl_intvec,
1082 .extra1 = &zero,
1085 .ctl_name = VM_MIN_UNMAPPED,
1086 .procname = "min_unmapped_ratio",
1087 .data = &sysctl_min_unmapped_ratio,
1088 .maxlen = sizeof(sysctl_min_unmapped_ratio),
1089 .mode = 0644,
1090 .proc_handler = &sysctl_min_unmapped_ratio_sysctl_handler,
1091 .strategy = &sysctl_intvec,
1092 .extra1 = &zero,
1093 .extra2 = &one_hundred,
1096 .ctl_name = VM_MIN_SLAB,
1097 .procname = "min_slab_ratio",
1098 .data = &sysctl_min_slab_ratio,
1099 .maxlen = sizeof(sysctl_min_slab_ratio),
1100 .mode = 0644,
1101 .proc_handler = &sysctl_min_slab_ratio_sysctl_handler,
1102 .strategy = &sysctl_intvec,
1103 .extra1 = &zero,
1104 .extra2 = &one_hundred,
1106 #endif
1107 #ifdef CONFIG_SMP
1109 .ctl_name = CTL_UNNUMBERED,
1110 .procname = "stat_interval",
1111 .data = &sysctl_stat_interval,
1112 .maxlen = sizeof(sysctl_stat_interval),
1113 .mode = 0644,
1114 .proc_handler = &proc_dointvec_jiffies,
1115 .strategy = &sysctl_jiffies,
1117 #endif
1118 #ifdef CONFIG_SECURITY
1120 .ctl_name = CTL_UNNUMBERED,
1121 .procname = "mmap_min_addr",
1122 .data = &mmap_min_addr,
1123 .maxlen = sizeof(unsigned long),
1124 .mode = 0644,
1125 .proc_handler = &proc_doulongvec_minmax,
1127 #endif
1128 #ifdef CONFIG_NUMA
1130 .ctl_name = CTL_UNNUMBERED,
1131 .procname = "numa_zonelist_order",
1132 .data = &numa_zonelist_order,
1133 .maxlen = NUMA_ZONELIST_ORDER_LEN,
1134 .mode = 0644,
1135 .proc_handler = &numa_zonelist_order_handler,
1136 .strategy = &sysctl_string,
1138 #endif
1139 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1140 (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1142 .ctl_name = VM_VDSO_ENABLED,
1143 .procname = "vdso_enabled",
1144 .data = &vdso_enabled,
1145 .maxlen = sizeof(vdso_enabled),
1146 .mode = 0644,
1147 .proc_handler = &proc_dointvec,
1148 .strategy = &sysctl_intvec,
1149 .extra1 = &zero,
1151 #endif
1152 #ifdef CONFIG_HIGHMEM
1154 .ctl_name = CTL_UNNUMBERED,
1155 .procname = "highmem_is_dirtyable",
1156 .data = &vm_highmem_is_dirtyable,
1157 .maxlen = sizeof(vm_highmem_is_dirtyable),
1158 .mode = 0644,
1159 .proc_handler = &proc_dointvec_minmax,
1160 .strategy = &sysctl_intvec,
1161 .extra1 = &zero,
1162 .extra2 = &one,
1164 #endif
1166 * NOTE: do not add new entries to this table unless you have read
1167 * Documentation/sysctl/ctl_unnumbered.txt
1169 { .ctl_name = 0 }
1172 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1173 static struct ctl_table binfmt_misc_table[] = {
1174 { .ctl_name = 0 }
1176 #endif
1178 static struct ctl_table fs_table[] = {
1180 .ctl_name = FS_NRINODE,
1181 .procname = "inode-nr",
1182 .data = &inodes_stat,
1183 .maxlen = 2*sizeof(int),
1184 .mode = 0444,
1185 .proc_handler = &proc_dointvec,
1188 .ctl_name = FS_STATINODE,
1189 .procname = "inode-state",
1190 .data = &inodes_stat,
1191 .maxlen = 7*sizeof(int),
1192 .mode = 0444,
1193 .proc_handler = &proc_dointvec,
1196 .procname = "file-nr",
1197 .data = &files_stat,
1198 .maxlen = 3*sizeof(int),
1199 .mode = 0444,
1200 .proc_handler = &proc_nr_files,
1203 .ctl_name = FS_MAXFILE,
1204 .procname = "file-max",
1205 .data = &files_stat.max_files,
1206 .maxlen = sizeof(int),
1207 .mode = 0644,
1208 .proc_handler = &proc_dointvec,
1211 .ctl_name = CTL_UNNUMBERED,
1212 .procname = "nr_open",
1213 .data = &sysctl_nr_open,
1214 .maxlen = sizeof(int),
1215 .mode = 0644,
1216 .proc_handler = &proc_dointvec_minmax,
1217 .extra1 = &sysctl_nr_open_min,
1218 .extra2 = &sysctl_nr_open_max,
1221 .ctl_name = FS_DENTRY,
1222 .procname = "dentry-state",
1223 .data = &dentry_stat,
1224 .maxlen = 6*sizeof(int),
1225 .mode = 0444,
1226 .proc_handler = &proc_dointvec,
1229 .ctl_name = FS_OVERFLOWUID,
1230 .procname = "overflowuid",
1231 .data = &fs_overflowuid,
1232 .maxlen = sizeof(int),
1233 .mode = 0644,
1234 .proc_handler = &proc_dointvec_minmax,
1235 .strategy = &sysctl_intvec,
1236 .extra1 = &minolduid,
1237 .extra2 = &maxolduid,
1240 .ctl_name = FS_OVERFLOWGID,
1241 .procname = "overflowgid",
1242 .data = &fs_overflowgid,
1243 .maxlen = sizeof(int),
1244 .mode = 0644,
1245 .proc_handler = &proc_dointvec_minmax,
1246 .strategy = &sysctl_intvec,
1247 .extra1 = &minolduid,
1248 .extra2 = &maxolduid,
1250 #ifdef CONFIG_FILE_LOCKING
1252 .ctl_name = FS_LEASES,
1253 .procname = "leases-enable",
1254 .data = &leases_enable,
1255 .maxlen = sizeof(int),
1256 .mode = 0644,
1257 .proc_handler = &proc_dointvec,
1259 #endif
1260 #ifdef CONFIG_DNOTIFY
1262 .ctl_name = FS_DIR_NOTIFY,
1263 .procname = "dir-notify-enable",
1264 .data = &dir_notify_enable,
1265 .maxlen = sizeof(int),
1266 .mode = 0644,
1267 .proc_handler = &proc_dointvec,
1269 #endif
1270 #ifdef CONFIG_MMU
1271 #ifdef CONFIG_FILE_LOCKING
1273 .ctl_name = FS_LEASE_TIME,
1274 .procname = "lease-break-time",
1275 .data = &lease_break_time,
1276 .maxlen = sizeof(int),
1277 .mode = 0644,
1278 .proc_handler = &proc_dointvec_minmax,
1279 .strategy = &sysctl_intvec,
1280 .extra1 = &zero,
1281 .extra2 = &two,
1283 #endif
1285 .procname = "aio-nr",
1286 .data = &aio_nr,
1287 .maxlen = sizeof(aio_nr),
1288 .mode = 0444,
1289 .proc_handler = &proc_doulongvec_minmax,
1292 .procname = "aio-max-nr",
1293 .data = &aio_max_nr,
1294 .maxlen = sizeof(aio_max_nr),
1295 .mode = 0644,
1296 .proc_handler = &proc_doulongvec_minmax,
1298 #ifdef CONFIG_INOTIFY_USER
1300 .ctl_name = FS_INOTIFY,
1301 .procname = "inotify",
1302 .mode = 0555,
1303 .child = inotify_table,
1305 #endif
1306 #endif
1308 .ctl_name = KERN_SETUID_DUMPABLE,
1309 .procname = "suid_dumpable",
1310 .data = &suid_dumpable,
1311 .maxlen = sizeof(int),
1312 .mode = 0644,
1313 .proc_handler = &proc_dointvec,
1315 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1317 .ctl_name = CTL_UNNUMBERED,
1318 .procname = "binfmt_misc",
1319 .mode = 0555,
1320 .child = binfmt_misc_table,
1322 #endif
1324 * NOTE: do not add new entries to this table unless you have read
1325 * Documentation/sysctl/ctl_unnumbered.txt
1327 { .ctl_name = 0 }
1330 static struct ctl_table debug_table[] = {
1331 #if defined(CONFIG_X86) || defined(CONFIG_PPC)
1333 .ctl_name = CTL_UNNUMBERED,
1334 .procname = "exception-trace",
1335 .data = &show_unhandled_signals,
1336 .maxlen = sizeof(int),
1337 .mode = 0644,
1338 .proc_handler = proc_dointvec
1340 #endif
1341 { .ctl_name = 0 }
1344 static struct ctl_table dev_table[] = {
1345 { .ctl_name = 0 }
1348 static DEFINE_SPINLOCK(sysctl_lock);
1350 /* called under sysctl_lock */
1351 static int use_table(struct ctl_table_header *p)
1353 if (unlikely(p->unregistering))
1354 return 0;
1355 p->used++;
1356 return 1;
1359 /* called under sysctl_lock */
1360 static void unuse_table(struct ctl_table_header *p)
1362 if (!--p->used)
1363 if (unlikely(p->unregistering))
1364 complete(p->unregistering);
1367 /* called under sysctl_lock, will reacquire if has to wait */
1368 static void start_unregistering(struct ctl_table_header *p)
1371 * if p->used is 0, nobody will ever touch that entry again;
1372 * we'll eliminate all paths to it before dropping sysctl_lock
1374 if (unlikely(p->used)) {
1375 struct completion wait;
1376 init_completion(&wait);
1377 p->unregistering = &wait;
1378 spin_unlock(&sysctl_lock);
1379 wait_for_completion(&wait);
1380 spin_lock(&sysctl_lock);
1381 } else {
1382 /* anything non-NULL; we'll never dereference it */
1383 p->unregistering = ERR_PTR(-EINVAL);
1386 * do not remove from the list until nobody holds it; walking the
1387 * list in do_sysctl() relies on that.
1389 list_del_init(&p->ctl_entry);
1392 void sysctl_head_get(struct ctl_table_header *head)
1394 spin_lock(&sysctl_lock);
1395 head->count++;
1396 spin_unlock(&sysctl_lock);
1399 void sysctl_head_put(struct ctl_table_header *head)
1401 spin_lock(&sysctl_lock);
1402 if (!--head->count)
1403 kfree(head);
1404 spin_unlock(&sysctl_lock);
1407 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1409 if (!head)
1410 BUG();
1411 spin_lock(&sysctl_lock);
1412 if (!use_table(head))
1413 head = ERR_PTR(-ENOENT);
1414 spin_unlock(&sysctl_lock);
1415 return head;
1418 void sysctl_head_finish(struct ctl_table_header *head)
1420 if (!head)
1421 return;
1422 spin_lock(&sysctl_lock);
1423 unuse_table(head);
1424 spin_unlock(&sysctl_lock);
1427 static struct ctl_table_set *
1428 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1430 struct ctl_table_set *set = &root->default_set;
1431 if (root->lookup)
1432 set = root->lookup(root, namespaces);
1433 return set;
1436 static struct list_head *
1437 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1439 struct ctl_table_set *set = lookup_header_set(root, namespaces);
1440 return &set->list;
1443 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1444 struct ctl_table_header *prev)
1446 struct ctl_table_root *root;
1447 struct list_head *header_list;
1448 struct ctl_table_header *head;
1449 struct list_head *tmp;
1451 spin_lock(&sysctl_lock);
1452 if (prev) {
1453 head = prev;
1454 tmp = &prev->ctl_entry;
1455 unuse_table(prev);
1456 goto next;
1458 tmp = &root_table_header.ctl_entry;
1459 for (;;) {
1460 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1462 if (!use_table(head))
1463 goto next;
1464 spin_unlock(&sysctl_lock);
1465 return head;
1466 next:
1467 root = head->root;
1468 tmp = tmp->next;
1469 header_list = lookup_header_list(root, namespaces);
1470 if (tmp != header_list)
1471 continue;
1473 do {
1474 root = list_entry(root->root_list.next,
1475 struct ctl_table_root, root_list);
1476 if (root == &sysctl_table_root)
1477 goto out;
1478 header_list = lookup_header_list(root, namespaces);
1479 } while (list_empty(header_list));
1480 tmp = header_list->next;
1482 out:
1483 spin_unlock(&sysctl_lock);
1484 return NULL;
1487 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1489 return __sysctl_head_next(current->nsproxy, prev);
1492 void register_sysctl_root(struct ctl_table_root *root)
1494 spin_lock(&sysctl_lock);
1495 list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1496 spin_unlock(&sysctl_lock);
1499 #ifdef CONFIG_SYSCTL_SYSCALL
1500 /* Perform the actual read/write of a sysctl table entry. */
1501 static int do_sysctl_strategy(struct ctl_table_root *root,
1502 struct ctl_table *table,
1503 void __user *oldval, size_t __user *oldlenp,
1504 void __user *newval, size_t newlen)
1506 int op = 0, rc;
1508 if (oldval)
1509 op |= MAY_READ;
1510 if (newval)
1511 op |= MAY_WRITE;
1512 if (sysctl_perm(root, table, op))
1513 return -EPERM;
1515 if (table->strategy) {
1516 rc = table->strategy(table, oldval, oldlenp, newval, newlen);
1517 if (rc < 0)
1518 return rc;
1519 if (rc > 0)
1520 return 0;
1523 /* If there is no strategy routine, or if the strategy returns
1524 * zero, proceed with automatic r/w */
1525 if (table->data && table->maxlen) {
1526 rc = sysctl_data(table, oldval, oldlenp, newval, newlen);
1527 if (rc < 0)
1528 return rc;
1530 return 0;
1533 static int parse_table(int __user *name, int nlen,
1534 void __user *oldval, size_t __user *oldlenp,
1535 void __user *newval, size_t newlen,
1536 struct ctl_table_root *root,
1537 struct ctl_table *table)
1539 int n;
1540 repeat:
1541 if (!nlen)
1542 return -ENOTDIR;
1543 if (get_user(n, name))
1544 return -EFAULT;
1545 for ( ; table->ctl_name || table->procname; table++) {
1546 if (!table->ctl_name)
1547 continue;
1548 if (n == table->ctl_name) {
1549 int error;
1550 if (table->child) {
1551 if (sysctl_perm(root, table, MAY_EXEC))
1552 return -EPERM;
1553 name++;
1554 nlen--;
1555 table = table->child;
1556 goto repeat;
1558 error = do_sysctl_strategy(root, table,
1559 oldval, oldlenp,
1560 newval, newlen);
1561 return error;
1564 return -ENOTDIR;
1567 int do_sysctl(int __user *name, int nlen, void __user *oldval, size_t __user *oldlenp,
1568 void __user *newval, size_t newlen)
1570 struct ctl_table_header *head;
1571 int error = -ENOTDIR;
1573 if (nlen <= 0 || nlen >= CTL_MAXNAME)
1574 return -ENOTDIR;
1575 if (oldval) {
1576 int old_len;
1577 if (!oldlenp || get_user(old_len, oldlenp))
1578 return -EFAULT;
1581 for (head = sysctl_head_next(NULL); head;
1582 head = sysctl_head_next(head)) {
1583 error = parse_table(name, nlen, oldval, oldlenp,
1584 newval, newlen,
1585 head->root, head->ctl_table);
1586 if (error != -ENOTDIR) {
1587 sysctl_head_finish(head);
1588 break;
1591 return error;
1594 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
1596 struct __sysctl_args tmp;
1597 int error;
1599 if (copy_from_user(&tmp, args, sizeof(tmp)))
1600 return -EFAULT;
1602 error = deprecated_sysctl_warning(&tmp);
1603 if (error)
1604 goto out;
1606 lock_kernel();
1607 error = do_sysctl(tmp.name, tmp.nlen, tmp.oldval, tmp.oldlenp,
1608 tmp.newval, tmp.newlen);
1609 unlock_kernel();
1610 out:
1611 return error;
1613 #endif /* CONFIG_SYSCTL_SYSCALL */
1616 * sysctl_perm does NOT grant the superuser all rights automatically, because
1617 * some sysctl variables are readonly even to root.
1620 static int test_perm(int mode, int op)
1622 if (!current->euid)
1623 mode >>= 6;
1624 else if (in_egroup_p(0))
1625 mode >>= 3;
1626 if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1627 return 0;
1628 return -EACCES;
1631 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1633 int error;
1634 int mode;
1636 error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1637 if (error)
1638 return error;
1640 if (root->permissions)
1641 mode = root->permissions(root, current->nsproxy, table);
1642 else
1643 mode = table->mode;
1645 return test_perm(mode, op);
1648 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1650 for (; table->ctl_name || table->procname; table++) {
1651 table->parent = parent;
1652 if (table->child)
1653 sysctl_set_parent(table, table->child);
1657 static __init int sysctl_init(void)
1659 sysctl_set_parent(NULL, root_table);
1660 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1662 int err;
1663 err = sysctl_check_table(current->nsproxy, root_table);
1665 #endif
1666 return 0;
1669 core_initcall(sysctl_init);
1671 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1672 struct ctl_table *table)
1674 struct ctl_table *p;
1675 const char *s = branch->procname;
1677 /* branch should have named subdirectory as its first element */
1678 if (!s || !branch->child)
1679 return NULL;
1681 /* ... and nothing else */
1682 if (branch[1].procname || branch[1].ctl_name)
1683 return NULL;
1685 /* table should contain subdirectory with the same name */
1686 for (p = table; p->procname || p->ctl_name; p++) {
1687 if (!p->child)
1688 continue;
1689 if (p->procname && strcmp(p->procname, s) == 0)
1690 return p;
1692 return NULL;
1695 /* see if attaching q to p would be an improvement */
1696 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1698 struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1699 struct ctl_table *next;
1700 int is_better = 0;
1701 int not_in_parent = !p->attached_by;
1703 while ((next = is_branch_in(by, to)) != NULL) {
1704 if (by == q->attached_by)
1705 is_better = 1;
1706 if (to == p->attached_by)
1707 not_in_parent = 1;
1708 by = by->child;
1709 to = next->child;
1712 if (is_better && not_in_parent) {
1713 q->attached_by = by;
1714 q->attached_to = to;
1715 q->parent = p;
1720 * __register_sysctl_paths - register a sysctl hierarchy
1721 * @root: List of sysctl headers to register on
1722 * @namespaces: Data to compute which lists of sysctl entries are visible
1723 * @path: The path to the directory the sysctl table is in.
1724 * @table: the top-level table structure
1726 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1727 * array. A completely 0 filled entry terminates the table.
1729 * The members of the &struct ctl_table structure are used as follows:
1731 * ctl_name - This is the numeric sysctl value used by sysctl(2). The number
1732 * must be unique within that level of sysctl
1734 * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1735 * enter a sysctl file
1737 * data - a pointer to data for use by proc_handler
1739 * maxlen - the maximum size in bytes of the data
1741 * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1743 * child - a pointer to the child sysctl table if this entry is a directory, or
1744 * %NULL.
1746 * proc_handler - the text handler routine (described below)
1748 * strategy - the strategy routine (described below)
1750 * de - for internal use by the sysctl routines
1752 * extra1, extra2 - extra pointers usable by the proc handler routines
1754 * Leaf nodes in the sysctl tree will be represented by a single file
1755 * under /proc; non-leaf nodes will be represented by directories.
1757 * sysctl(2) can automatically manage read and write requests through
1758 * the sysctl table. The data and maxlen fields of the ctl_table
1759 * struct enable minimal validation of the values being written to be
1760 * performed, and the mode field allows minimal authentication.
1762 * More sophisticated management can be enabled by the provision of a
1763 * strategy routine with the table entry. This will be called before
1764 * any automatic read or write of the data is performed.
1766 * The strategy routine may return
1768 * < 0 - Error occurred (error is passed to user process)
1770 * 0 - OK - proceed with automatic read or write.
1772 * > 0 - OK - read or write has been done by the strategy routine, so
1773 * return immediately.
1775 * There must be a proc_handler routine for any terminal nodes
1776 * mirrored under /proc/sys (non-terminals are handled by a built-in
1777 * directory handler). Several default handlers are available to
1778 * cover common cases -
1780 * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1781 * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1782 * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1784 * It is the handler's job to read the input buffer from user memory
1785 * and process it. The handler should return 0 on success.
1787 * This routine returns %NULL on a failure to register, and a pointer
1788 * to the table header on success.
1790 struct ctl_table_header *__register_sysctl_paths(
1791 struct ctl_table_root *root,
1792 struct nsproxy *namespaces,
1793 const struct ctl_path *path, struct ctl_table *table)
1795 struct ctl_table_header *header;
1796 struct ctl_table *new, **prevp;
1797 unsigned int n, npath;
1798 struct ctl_table_set *set;
1800 /* Count the path components */
1801 for (npath = 0; path[npath].ctl_name || path[npath].procname; ++npath)
1805 * For each path component, allocate a 2-element ctl_table array.
1806 * The first array element will be filled with the sysctl entry
1807 * for this, the second will be the sentinel (ctl_name == 0).
1809 * We allocate everything in one go so that we don't have to
1810 * worry about freeing additional memory in unregister_sysctl_table.
1812 header = kzalloc(sizeof(struct ctl_table_header) +
1813 (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1814 if (!header)
1815 return NULL;
1817 new = (struct ctl_table *) (header + 1);
1819 /* Now connect the dots */
1820 prevp = &header->ctl_table;
1821 for (n = 0; n < npath; ++n, ++path) {
1822 /* Copy the procname */
1823 new->procname = path->procname;
1824 new->ctl_name = path->ctl_name;
1825 new->mode = 0555;
1827 *prevp = new;
1828 prevp = &new->child;
1830 new += 2;
1832 *prevp = table;
1833 header->ctl_table_arg = table;
1835 INIT_LIST_HEAD(&header->ctl_entry);
1836 header->used = 0;
1837 header->unregistering = NULL;
1838 header->root = root;
1839 sysctl_set_parent(NULL, header->ctl_table);
1840 header->count = 1;
1841 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1842 if (sysctl_check_table(namespaces, header->ctl_table)) {
1843 kfree(header);
1844 return NULL;
1846 #endif
1847 spin_lock(&sysctl_lock);
1848 header->set = lookup_header_set(root, namespaces);
1849 header->attached_by = header->ctl_table;
1850 header->attached_to = root_table;
1851 header->parent = &root_table_header;
1852 for (set = header->set; set; set = set->parent) {
1853 struct ctl_table_header *p;
1854 list_for_each_entry(p, &set->list, ctl_entry) {
1855 if (p->unregistering)
1856 continue;
1857 try_attach(p, header);
1860 header->parent->count++;
1861 list_add_tail(&header->ctl_entry, &header->set->list);
1862 spin_unlock(&sysctl_lock);
1864 return header;
1868 * register_sysctl_table_path - register a sysctl table hierarchy
1869 * @path: The path to the directory the sysctl table is in.
1870 * @table: the top-level table structure
1872 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1873 * array. A completely 0 filled entry terminates the table.
1875 * See __register_sysctl_paths for more details.
1877 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1878 struct ctl_table *table)
1880 return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1881 path, table);
1885 * register_sysctl_table - register a sysctl table hierarchy
1886 * @table: the top-level table structure
1888 * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1889 * array. A completely 0 filled entry terminates the table.
1891 * See register_sysctl_paths for more details.
1893 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1895 static const struct ctl_path null_path[] = { {} };
1897 return register_sysctl_paths(null_path, table);
1901 * unregister_sysctl_table - unregister a sysctl table hierarchy
1902 * @header: the header returned from register_sysctl_table
1904 * Unregisters the sysctl table and all children. proc entries may not
1905 * actually be removed until they are no longer used by anyone.
1907 void unregister_sysctl_table(struct ctl_table_header * header)
1909 might_sleep();
1911 if (header == NULL)
1912 return;
1914 spin_lock(&sysctl_lock);
1915 start_unregistering(header);
1916 if (!--header->parent->count) {
1917 WARN_ON(1);
1918 kfree(header->parent);
1920 if (!--header->count)
1921 kfree(header);
1922 spin_unlock(&sysctl_lock);
1925 int sysctl_is_seen(struct ctl_table_header *p)
1927 struct ctl_table_set *set = p->set;
1928 int res;
1929 spin_lock(&sysctl_lock);
1930 if (p->unregistering)
1931 res = 0;
1932 else if (!set->is_seen)
1933 res = 1;
1934 else
1935 res = set->is_seen(set);
1936 spin_unlock(&sysctl_lock);
1937 return res;
1940 void setup_sysctl_set(struct ctl_table_set *p,
1941 struct ctl_table_set *parent,
1942 int (*is_seen)(struct ctl_table_set *))
1944 INIT_LIST_HEAD(&p->list);
1945 p->parent = parent ? parent : &sysctl_table_root.default_set;
1946 p->is_seen = is_seen;
1949 #else /* !CONFIG_SYSCTL */
1950 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1952 return NULL;
1955 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1956 struct ctl_table *table)
1958 return NULL;
1961 void unregister_sysctl_table(struct ctl_table_header * table)
1965 void setup_sysctl_set(struct ctl_table_set *p,
1966 struct ctl_table_set *parent,
1967 int (*is_seen)(struct ctl_table_set *))
1971 void sysctl_head_put(struct ctl_table_header *head)
1975 #endif /* CONFIG_SYSCTL */
1978 * /proc/sys support
1981 #ifdef CONFIG_PROC_SYSCTL
1983 static int _proc_do_string(void* data, int maxlen, int write,
1984 struct file *filp, void __user *buffer,
1985 size_t *lenp, loff_t *ppos)
1987 size_t len;
1988 char __user *p;
1989 char c;
1991 if (!data || !maxlen || !*lenp) {
1992 *lenp = 0;
1993 return 0;
1996 if (write) {
1997 len = 0;
1998 p = buffer;
1999 while (len < *lenp) {
2000 if (get_user(c, p++))
2001 return -EFAULT;
2002 if (c == 0 || c == '\n')
2003 break;
2004 len++;
2006 if (len >= maxlen)
2007 len = maxlen-1;
2008 if(copy_from_user(data, buffer, len))
2009 return -EFAULT;
2010 ((char *) data)[len] = 0;
2011 *ppos += *lenp;
2012 } else {
2013 len = strlen(data);
2014 if (len > maxlen)
2015 len = maxlen;
2017 if (*ppos > len) {
2018 *lenp = 0;
2019 return 0;
2022 data += *ppos;
2023 len -= *ppos;
2025 if (len > *lenp)
2026 len = *lenp;
2027 if (len)
2028 if(copy_to_user(buffer, data, len))
2029 return -EFAULT;
2030 if (len < *lenp) {
2031 if(put_user('\n', ((char __user *) buffer) + len))
2032 return -EFAULT;
2033 len++;
2035 *lenp = len;
2036 *ppos += len;
2038 return 0;
2042 * proc_dostring - read a string sysctl
2043 * @table: the sysctl table
2044 * @write: %TRUE if this is a write to the sysctl file
2045 * @filp: the file structure
2046 * @buffer: the user buffer
2047 * @lenp: the size of the user buffer
2048 * @ppos: file position
2050 * Reads/writes a string from/to the user buffer. If the kernel
2051 * buffer provided is not large enough to hold the string, the
2052 * string is truncated. The copied string is %NULL-terminated.
2053 * If the string is being read by the user process, it is copied
2054 * and a newline '\n' is added. It is truncated if the buffer is
2055 * not large enough.
2057 * Returns 0 on success.
2059 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
2060 void __user *buffer, size_t *lenp, loff_t *ppos)
2062 return _proc_do_string(table->data, table->maxlen, write, filp,
2063 buffer, lenp, ppos);
2067 static int do_proc_dointvec_conv(int *negp, unsigned long *lvalp,
2068 int *valp,
2069 int write, void *data)
2071 if (write) {
2072 *valp = *negp ? -*lvalp : *lvalp;
2073 } else {
2074 int val = *valp;
2075 if (val < 0) {
2076 *negp = -1;
2077 *lvalp = (unsigned long)-val;
2078 } else {
2079 *negp = 0;
2080 *lvalp = (unsigned long)val;
2083 return 0;
2086 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2087 int write, struct file *filp, void __user *buffer,
2088 size_t *lenp, loff_t *ppos,
2089 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2090 int write, void *data),
2091 void *data)
2093 #define TMPBUFLEN 21
2094 int *i, vleft, first=1, neg, val;
2095 unsigned long lval;
2096 size_t left, len;
2098 char buf[TMPBUFLEN], *p;
2099 char __user *s = buffer;
2101 if (!tbl_data || !table->maxlen || !*lenp ||
2102 (*ppos && !write)) {
2103 *lenp = 0;
2104 return 0;
2107 i = (int *) tbl_data;
2108 vleft = table->maxlen / sizeof(*i);
2109 left = *lenp;
2111 if (!conv)
2112 conv = do_proc_dointvec_conv;
2114 for (; left && vleft--; i++, first=0) {
2115 if (write) {
2116 while (left) {
2117 char c;
2118 if (get_user(c, s))
2119 return -EFAULT;
2120 if (!isspace(c))
2121 break;
2122 left--;
2123 s++;
2125 if (!left)
2126 break;
2127 neg = 0;
2128 len = left;
2129 if (len > sizeof(buf) - 1)
2130 len = sizeof(buf) - 1;
2131 if (copy_from_user(buf, s, len))
2132 return -EFAULT;
2133 buf[len] = 0;
2134 p = buf;
2135 if (*p == '-' && left > 1) {
2136 neg = 1;
2137 p++;
2139 if (*p < '0' || *p > '9')
2140 break;
2142 lval = simple_strtoul(p, &p, 0);
2144 len = p-buf;
2145 if ((len < left) && *p && !isspace(*p))
2146 break;
2147 if (neg)
2148 val = -val;
2149 s += len;
2150 left -= len;
2152 if (conv(&neg, &lval, i, 1, data))
2153 break;
2154 } else {
2155 p = buf;
2156 if (!first)
2157 *p++ = '\t';
2159 if (conv(&neg, &lval, i, 0, data))
2160 break;
2162 sprintf(p, "%s%lu", neg ? "-" : "", lval);
2163 len = strlen(buf);
2164 if (len > left)
2165 len = left;
2166 if(copy_to_user(s, buf, len))
2167 return -EFAULT;
2168 left -= len;
2169 s += len;
2173 if (!write && !first && left) {
2174 if(put_user('\n', s))
2175 return -EFAULT;
2176 left--, s++;
2178 if (write) {
2179 while (left) {
2180 char c;
2181 if (get_user(c, s++))
2182 return -EFAULT;
2183 if (!isspace(c))
2184 break;
2185 left--;
2188 if (write && first)
2189 return -EINVAL;
2190 *lenp -= left;
2191 *ppos += *lenp;
2192 return 0;
2193 #undef TMPBUFLEN
2196 static int do_proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2197 void __user *buffer, size_t *lenp, loff_t *ppos,
2198 int (*conv)(int *negp, unsigned long *lvalp, int *valp,
2199 int write, void *data),
2200 void *data)
2202 return __do_proc_dointvec(table->data, table, write, filp,
2203 buffer, lenp, ppos, conv, data);
2207 * proc_dointvec - read a vector of integers
2208 * @table: the sysctl table
2209 * @write: %TRUE if this is a write to the sysctl file
2210 * @filp: the file structure
2211 * @buffer: the user buffer
2212 * @lenp: the size of the user buffer
2213 * @ppos: file position
2215 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2216 * values from/to the user buffer, treated as an ASCII string.
2218 * Returns 0 on success.
2220 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2221 void __user *buffer, size_t *lenp, loff_t *ppos)
2223 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2224 NULL,NULL);
2228 * Taint values can only be increased
2229 * This means we can safely use a temporary.
2231 static int proc_taint(struct ctl_table *table, int write, struct file *filp,
2232 void __user *buffer, size_t *lenp, loff_t *ppos)
2234 struct ctl_table t;
2235 unsigned long tmptaint = get_taint();
2236 int err;
2238 if (write && !capable(CAP_SYS_ADMIN))
2239 return -EPERM;
2241 t = *table;
2242 t.data = &tmptaint;
2243 err = proc_doulongvec_minmax(&t, write, filp, buffer, lenp, ppos);
2244 if (err < 0)
2245 return err;
2247 if (write) {
2249 * Poor man's atomic or. Not worth adding a primitive
2250 * to everyone's atomic.h for this
2252 int i;
2253 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2254 if ((tmptaint >> i) & 1)
2255 add_taint(i);
2259 return err;
2262 struct do_proc_dointvec_minmax_conv_param {
2263 int *min;
2264 int *max;
2267 static int do_proc_dointvec_minmax_conv(int *negp, unsigned long *lvalp,
2268 int *valp,
2269 int write, void *data)
2271 struct do_proc_dointvec_minmax_conv_param *param = data;
2272 if (write) {
2273 int val = *negp ? -*lvalp : *lvalp;
2274 if ((param->min && *param->min > val) ||
2275 (param->max && *param->max < val))
2276 return -EINVAL;
2277 *valp = val;
2278 } else {
2279 int val = *valp;
2280 if (val < 0) {
2281 *negp = -1;
2282 *lvalp = (unsigned long)-val;
2283 } else {
2284 *negp = 0;
2285 *lvalp = (unsigned long)val;
2288 return 0;
2292 * proc_dointvec_minmax - read a vector of integers with min/max values
2293 * @table: the sysctl table
2294 * @write: %TRUE if this is a write to the sysctl file
2295 * @filp: the file structure
2296 * @buffer: the user buffer
2297 * @lenp: the size of the user buffer
2298 * @ppos: file position
2300 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2301 * values from/to the user buffer, treated as an ASCII string.
2303 * This routine will ensure the values are within the range specified by
2304 * table->extra1 (min) and table->extra2 (max).
2306 * Returns 0 on success.
2308 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
2309 void __user *buffer, size_t *lenp, loff_t *ppos)
2311 struct do_proc_dointvec_minmax_conv_param param = {
2312 .min = (int *) table->extra1,
2313 .max = (int *) table->extra2,
2315 return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2316 do_proc_dointvec_minmax_conv, &param);
2319 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2320 struct file *filp,
2321 void __user *buffer,
2322 size_t *lenp, loff_t *ppos,
2323 unsigned long convmul,
2324 unsigned long convdiv)
2326 #define TMPBUFLEN 21
2327 unsigned long *i, *min, *max, val;
2328 int vleft, first=1, neg;
2329 size_t len, left;
2330 char buf[TMPBUFLEN], *p;
2331 char __user *s = buffer;
2333 if (!data || !table->maxlen || !*lenp ||
2334 (*ppos && !write)) {
2335 *lenp = 0;
2336 return 0;
2339 i = (unsigned long *) data;
2340 min = (unsigned long *) table->extra1;
2341 max = (unsigned long *) table->extra2;
2342 vleft = table->maxlen / sizeof(unsigned long);
2343 left = *lenp;
2345 for (; left && vleft--; i++, min++, max++, first=0) {
2346 if (write) {
2347 while (left) {
2348 char c;
2349 if (get_user(c, s))
2350 return -EFAULT;
2351 if (!isspace(c))
2352 break;
2353 left--;
2354 s++;
2356 if (!left)
2357 break;
2358 neg = 0;
2359 len = left;
2360 if (len > TMPBUFLEN-1)
2361 len = TMPBUFLEN-1;
2362 if (copy_from_user(buf, s, len))
2363 return -EFAULT;
2364 buf[len] = 0;
2365 p = buf;
2366 if (*p == '-' && left > 1) {
2367 neg = 1;
2368 p++;
2370 if (*p < '0' || *p > '9')
2371 break;
2372 val = simple_strtoul(p, &p, 0) * convmul / convdiv ;
2373 len = p-buf;
2374 if ((len < left) && *p && !isspace(*p))
2375 break;
2376 if (neg)
2377 val = -val;
2378 s += len;
2379 left -= len;
2381 if(neg)
2382 continue;
2383 if ((min && val < *min) || (max && val > *max))
2384 continue;
2385 *i = val;
2386 } else {
2387 p = buf;
2388 if (!first)
2389 *p++ = '\t';
2390 sprintf(p, "%lu", convdiv * (*i) / convmul);
2391 len = strlen(buf);
2392 if (len > left)
2393 len = left;
2394 if(copy_to_user(s, buf, len))
2395 return -EFAULT;
2396 left -= len;
2397 s += len;
2401 if (!write && !first && left) {
2402 if(put_user('\n', s))
2403 return -EFAULT;
2404 left--, s++;
2406 if (write) {
2407 while (left) {
2408 char c;
2409 if (get_user(c, s++))
2410 return -EFAULT;
2411 if (!isspace(c))
2412 break;
2413 left--;
2416 if (write && first)
2417 return -EINVAL;
2418 *lenp -= left;
2419 *ppos += *lenp;
2420 return 0;
2421 #undef TMPBUFLEN
2424 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2425 struct file *filp,
2426 void __user *buffer,
2427 size_t *lenp, loff_t *ppos,
2428 unsigned long convmul,
2429 unsigned long convdiv)
2431 return __do_proc_doulongvec_minmax(table->data, table, write,
2432 filp, buffer, lenp, ppos, convmul, convdiv);
2436 * proc_doulongvec_minmax - read a vector of long integers with min/max values
2437 * @table: the sysctl table
2438 * @write: %TRUE if this is a write to the sysctl file
2439 * @filp: the file structure
2440 * @buffer: the user buffer
2441 * @lenp: the size of the user buffer
2442 * @ppos: file position
2444 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2445 * values from/to the user buffer, treated as an ASCII string.
2447 * This routine will ensure the values are within the range specified by
2448 * table->extra1 (min) and table->extra2 (max).
2450 * Returns 0 on success.
2452 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2453 void __user *buffer, size_t *lenp, loff_t *ppos)
2455 return do_proc_doulongvec_minmax(table, write, filp, buffer, lenp, ppos, 1l, 1l);
2459 * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2460 * @table: the sysctl table
2461 * @write: %TRUE if this is a write to the sysctl file
2462 * @filp: the file structure
2463 * @buffer: the user buffer
2464 * @lenp: the size of the user buffer
2465 * @ppos: file position
2467 * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2468 * values from/to the user buffer, treated as an ASCII string. The values
2469 * are treated as milliseconds, and converted to jiffies when they are stored.
2471 * This routine will ensure the values are within the range specified by
2472 * table->extra1 (min) and table->extra2 (max).
2474 * Returns 0 on success.
2476 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2477 struct file *filp,
2478 void __user *buffer,
2479 size_t *lenp, loff_t *ppos)
2481 return do_proc_doulongvec_minmax(table, write, filp, buffer,
2482 lenp, ppos, HZ, 1000l);
2486 static int do_proc_dointvec_jiffies_conv(int *negp, unsigned long *lvalp,
2487 int *valp,
2488 int write, void *data)
2490 if (write) {
2491 if (*lvalp > LONG_MAX / HZ)
2492 return 1;
2493 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2494 } else {
2495 int val = *valp;
2496 unsigned long lval;
2497 if (val < 0) {
2498 *negp = -1;
2499 lval = (unsigned long)-val;
2500 } else {
2501 *negp = 0;
2502 lval = (unsigned long)val;
2504 *lvalp = lval / HZ;
2506 return 0;
2509 static int do_proc_dointvec_userhz_jiffies_conv(int *negp, unsigned long *lvalp,
2510 int *valp,
2511 int write, void *data)
2513 if (write) {
2514 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2515 return 1;
2516 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2517 } else {
2518 int val = *valp;
2519 unsigned long lval;
2520 if (val < 0) {
2521 *negp = -1;
2522 lval = (unsigned long)-val;
2523 } else {
2524 *negp = 0;
2525 lval = (unsigned long)val;
2527 *lvalp = jiffies_to_clock_t(lval);
2529 return 0;
2532 static int do_proc_dointvec_ms_jiffies_conv(int *negp, unsigned long *lvalp,
2533 int *valp,
2534 int write, void *data)
2536 if (write) {
2537 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2538 } else {
2539 int val = *valp;
2540 unsigned long lval;
2541 if (val < 0) {
2542 *negp = -1;
2543 lval = (unsigned long)-val;
2544 } else {
2545 *negp = 0;
2546 lval = (unsigned long)val;
2548 *lvalp = jiffies_to_msecs(lval);
2550 return 0;
2554 * proc_dointvec_jiffies - read a vector of integers as seconds
2555 * @table: the sysctl table
2556 * @write: %TRUE if this is a write to the sysctl file
2557 * @filp: the file structure
2558 * @buffer: the user buffer
2559 * @lenp: the size of the user buffer
2560 * @ppos: file position
2562 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2563 * values from/to the user buffer, treated as an ASCII string.
2564 * The values read are assumed to be in seconds, and are converted into
2565 * jiffies.
2567 * Returns 0 on success.
2569 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2570 void __user *buffer, size_t *lenp, loff_t *ppos)
2572 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2573 do_proc_dointvec_jiffies_conv,NULL);
2577 * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2578 * @table: the sysctl table
2579 * @write: %TRUE if this is a write to the sysctl file
2580 * @filp: the file structure
2581 * @buffer: the user buffer
2582 * @lenp: the size of the user buffer
2583 * @ppos: pointer to the file position
2585 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2586 * values from/to the user buffer, treated as an ASCII string.
2587 * The values read are assumed to be in 1/USER_HZ seconds, and
2588 * are converted into jiffies.
2590 * Returns 0 on success.
2592 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2593 void __user *buffer, size_t *lenp, loff_t *ppos)
2595 return do_proc_dointvec(table,write,filp,buffer,lenp,ppos,
2596 do_proc_dointvec_userhz_jiffies_conv,NULL);
2600 * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2601 * @table: the sysctl table
2602 * @write: %TRUE if this is a write to the sysctl file
2603 * @filp: the file structure
2604 * @buffer: the user buffer
2605 * @lenp: the size of the user buffer
2606 * @ppos: file position
2607 * @ppos: the current position in the file
2609 * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2610 * values from/to the user buffer, treated as an ASCII string.
2611 * The values read are assumed to be in 1/1000 seconds, and
2612 * are converted into jiffies.
2614 * Returns 0 on success.
2616 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2617 void __user *buffer, size_t *lenp, loff_t *ppos)
2619 return do_proc_dointvec(table, write, filp, buffer, lenp, ppos,
2620 do_proc_dointvec_ms_jiffies_conv, NULL);
2623 static int proc_do_cad_pid(struct ctl_table *table, int write, struct file *filp,
2624 void __user *buffer, size_t *lenp, loff_t *ppos)
2626 struct pid *new_pid;
2627 pid_t tmp;
2628 int r;
2630 tmp = pid_vnr(cad_pid);
2632 r = __do_proc_dointvec(&tmp, table, write, filp, buffer,
2633 lenp, ppos, NULL, NULL);
2634 if (r || !write)
2635 return r;
2637 new_pid = find_get_pid(tmp);
2638 if (!new_pid)
2639 return -ESRCH;
2641 put_pid(xchg(&cad_pid, new_pid));
2642 return 0;
2645 #else /* CONFIG_PROC_FS */
2647 int proc_dostring(struct ctl_table *table, int write, struct file *filp,
2648 void __user *buffer, size_t *lenp, loff_t *ppos)
2650 return -ENOSYS;
2653 int proc_dointvec(struct ctl_table *table, int write, struct file *filp,
2654 void __user *buffer, size_t *lenp, loff_t *ppos)
2656 return -ENOSYS;
2659 int proc_dointvec_minmax(struct ctl_table *table, int write, struct file *filp,
2660 void __user *buffer, size_t *lenp, loff_t *ppos)
2662 return -ENOSYS;
2665 int proc_dointvec_jiffies(struct ctl_table *table, int write, struct file *filp,
2666 void __user *buffer, size_t *lenp, loff_t *ppos)
2668 return -ENOSYS;
2671 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write, struct file *filp,
2672 void __user *buffer, size_t *lenp, loff_t *ppos)
2674 return -ENOSYS;
2677 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write, struct file *filp,
2678 void __user *buffer, size_t *lenp, loff_t *ppos)
2680 return -ENOSYS;
2683 int proc_doulongvec_minmax(struct ctl_table *table, int write, struct file *filp,
2684 void __user *buffer, size_t *lenp, loff_t *ppos)
2686 return -ENOSYS;
2689 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2690 struct file *filp,
2691 void __user *buffer,
2692 size_t *lenp, loff_t *ppos)
2694 return -ENOSYS;
2698 #endif /* CONFIG_PROC_FS */
2701 #ifdef CONFIG_SYSCTL_SYSCALL
2703 * General sysctl support routines
2706 /* The generic sysctl data routine (used if no strategy routine supplied) */
2707 int sysctl_data(struct ctl_table *table,
2708 void __user *oldval, size_t __user *oldlenp,
2709 void __user *newval, size_t newlen)
2711 size_t len;
2713 /* Get out of I don't have a variable */
2714 if (!table->data || !table->maxlen)
2715 return -ENOTDIR;
2717 if (oldval && oldlenp) {
2718 if (get_user(len, oldlenp))
2719 return -EFAULT;
2720 if (len) {
2721 if (len > table->maxlen)
2722 len = table->maxlen;
2723 if (copy_to_user(oldval, table->data, len))
2724 return -EFAULT;
2725 if (put_user(len, oldlenp))
2726 return -EFAULT;
2730 if (newval && newlen) {
2731 if (newlen > table->maxlen)
2732 newlen = table->maxlen;
2734 if (copy_from_user(table->data, newval, newlen))
2735 return -EFAULT;
2737 return 1;
2740 /* The generic string strategy routine: */
2741 int sysctl_string(struct ctl_table *table,
2742 void __user *oldval, size_t __user *oldlenp,
2743 void __user *newval, size_t newlen)
2745 if (!table->data || !table->maxlen)
2746 return -ENOTDIR;
2748 if (oldval && oldlenp) {
2749 size_t bufsize;
2750 if (get_user(bufsize, oldlenp))
2751 return -EFAULT;
2752 if (bufsize) {
2753 size_t len = strlen(table->data), copied;
2755 /* This shouldn't trigger for a well-formed sysctl */
2756 if (len > table->maxlen)
2757 len = table->maxlen;
2759 /* Copy up to a max of bufsize-1 bytes of the string */
2760 copied = (len >= bufsize) ? bufsize - 1 : len;
2762 if (copy_to_user(oldval, table->data, copied) ||
2763 put_user(0, (char __user *)(oldval + copied)))
2764 return -EFAULT;
2765 if (put_user(len, oldlenp))
2766 return -EFAULT;
2769 if (newval && newlen) {
2770 size_t len = newlen;
2771 if (len > table->maxlen)
2772 len = table->maxlen;
2773 if(copy_from_user(table->data, newval, len))
2774 return -EFAULT;
2775 if (len == table->maxlen)
2776 len--;
2777 ((char *) table->data)[len] = 0;
2779 return 1;
2783 * This function makes sure that all of the integers in the vector
2784 * are between the minimum and maximum values given in the arrays
2785 * table->extra1 and table->extra2, respectively.
2787 int sysctl_intvec(struct ctl_table *table,
2788 void __user *oldval, size_t __user *oldlenp,
2789 void __user *newval, size_t newlen)
2792 if (newval && newlen) {
2793 int __user *vec = (int __user *) newval;
2794 int *min = (int *) table->extra1;
2795 int *max = (int *) table->extra2;
2796 size_t length;
2797 int i;
2799 if (newlen % sizeof(int) != 0)
2800 return -EINVAL;
2802 if (!table->extra1 && !table->extra2)
2803 return 0;
2805 if (newlen > table->maxlen)
2806 newlen = table->maxlen;
2807 length = newlen / sizeof(int);
2809 for (i = 0; i < length; i++) {
2810 int value;
2811 if (get_user(value, vec + i))
2812 return -EFAULT;
2813 if (min && value < min[i])
2814 return -EINVAL;
2815 if (max && value > max[i])
2816 return -EINVAL;
2819 return 0;
2822 /* Strategy function to convert jiffies to seconds */
2823 int sysctl_jiffies(struct ctl_table *table,
2824 void __user *oldval, size_t __user *oldlenp,
2825 void __user *newval, size_t newlen)
2827 if (oldval && oldlenp) {
2828 size_t olen;
2830 if (get_user(olen, oldlenp))
2831 return -EFAULT;
2832 if (olen) {
2833 int val;
2835 if (olen < sizeof(int))
2836 return -EINVAL;
2838 val = *(int *)(table->data) / HZ;
2839 if (put_user(val, (int __user *)oldval))
2840 return -EFAULT;
2841 if (put_user(sizeof(int), oldlenp))
2842 return -EFAULT;
2845 if (newval && newlen) {
2846 int new;
2847 if (newlen != sizeof(int))
2848 return -EINVAL;
2849 if (get_user(new, (int __user *)newval))
2850 return -EFAULT;
2851 *(int *)(table->data) = new*HZ;
2853 return 1;
2856 /* Strategy function to convert jiffies to seconds */
2857 int sysctl_ms_jiffies(struct ctl_table *table,
2858 void __user *oldval, size_t __user *oldlenp,
2859 void __user *newval, size_t newlen)
2861 if (oldval && oldlenp) {
2862 size_t olen;
2864 if (get_user(olen, oldlenp))
2865 return -EFAULT;
2866 if (olen) {
2867 int val;
2869 if (olen < sizeof(int))
2870 return -EINVAL;
2872 val = jiffies_to_msecs(*(int *)(table->data));
2873 if (put_user(val, (int __user *)oldval))
2874 return -EFAULT;
2875 if (put_user(sizeof(int), oldlenp))
2876 return -EFAULT;
2879 if (newval && newlen) {
2880 int new;
2881 if (newlen != sizeof(int))
2882 return -EINVAL;
2883 if (get_user(new, (int __user *)newval))
2884 return -EFAULT;
2885 *(int *)(table->data) = msecs_to_jiffies(new);
2887 return 1;
2892 #else /* CONFIG_SYSCTL_SYSCALL */
2895 asmlinkage long sys_sysctl(struct __sysctl_args __user *args)
2897 struct __sysctl_args tmp;
2898 int error;
2900 if (copy_from_user(&tmp, args, sizeof(tmp)))
2901 return -EFAULT;
2903 error = deprecated_sysctl_warning(&tmp);
2905 /* If no error reading the parameters then just -ENOSYS ... */
2906 if (!error)
2907 error = -ENOSYS;
2909 return error;
2912 int sysctl_data(struct ctl_table *table,
2913 void __user *oldval, size_t __user *oldlenp,
2914 void __user *newval, size_t newlen)
2916 return -ENOSYS;
2919 int sysctl_string(struct ctl_table *table,
2920 void __user *oldval, size_t __user *oldlenp,
2921 void __user *newval, size_t newlen)
2923 return -ENOSYS;
2926 int sysctl_intvec(struct ctl_table *table,
2927 void __user *oldval, size_t __user *oldlenp,
2928 void __user *newval, size_t newlen)
2930 return -ENOSYS;
2933 int sysctl_jiffies(struct ctl_table *table,
2934 void __user *oldval, size_t __user *oldlenp,
2935 void __user *newval, size_t newlen)
2937 return -ENOSYS;
2940 int sysctl_ms_jiffies(struct ctl_table *table,
2941 void __user *oldval, size_t __user *oldlenp,
2942 void __user *newval, size_t newlen)
2944 return -ENOSYS;
2947 #endif /* CONFIG_SYSCTL_SYSCALL */
2949 static int deprecated_sysctl_warning(struct __sysctl_args *args)
2951 static int msg_count;
2952 int name[CTL_MAXNAME];
2953 int i;
2955 /* Check args->nlen. */
2956 if (args->nlen < 0 || args->nlen > CTL_MAXNAME)
2957 return -ENOTDIR;
2959 /* Read in the sysctl name for better debug message logging */
2960 for (i = 0; i < args->nlen; i++)
2961 if (get_user(name[i], args->name + i))
2962 return -EFAULT;
2964 /* Ignore accesses to kernel.version */
2965 if ((args->nlen == 2) && (name[0] == CTL_KERN) && (name[1] == KERN_VERSION))
2966 return 0;
2968 if (msg_count < 5) {
2969 msg_count++;
2970 printk(KERN_INFO
2971 "warning: process `%s' used the deprecated sysctl "
2972 "system call with ", current->comm);
2973 for (i = 0; i < args->nlen; i++)
2974 printk("%d.", name[i]);
2975 printk("\n");
2977 return 0;
2981 * No sense putting this after each symbol definition, twice,
2982 * exception granted :-)
2984 EXPORT_SYMBOL(proc_dointvec);
2985 EXPORT_SYMBOL(proc_dointvec_jiffies);
2986 EXPORT_SYMBOL(proc_dointvec_minmax);
2987 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2988 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2989 EXPORT_SYMBOL(proc_dostring);
2990 EXPORT_SYMBOL(proc_doulongvec_minmax);
2991 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2992 EXPORT_SYMBOL(register_sysctl_table);
2993 EXPORT_SYMBOL(register_sysctl_paths);
2994 EXPORT_SYMBOL(sysctl_intvec);
2995 EXPORT_SYMBOL(sysctl_jiffies);
2996 EXPORT_SYMBOL(sysctl_ms_jiffies);
2997 EXPORT_SYMBOL(sysctl_string);
2998 EXPORT_SYMBOL(sysctl_data);
2999 EXPORT_SYMBOL(unregister_sysctl_table);